Re: RFE: ncat should call shutdown on EOF

[Michal Hlavinka <mhlavink () redhat com>]

On 09/10/2012 10:17 PM, David Fifield wrote:
> On Mon, Sep 10, 2012 at 09:23:20AM +0200, Michal Hlavinka wrote:
> > On 09/07/2012 03:43 PM, Henri Doreau wrote:
> > > 2012/8/2 Michal Hlavinka <mhlavink () redhat com>:
> > > > Hi,
> > > >
> > > > we're replacing nc in fedora with ncat and testing if the behaviour is
> > > > similar enough.
> > > >
> > > > One of the differences is old nc calls shutdown on EOF. When ncat is used
> > > > and gets EOF (ctrl-d) as an input, it does nothing unless --send-only is
> > > > used. Would it be possible to call shutdown(fd,SHUT_WR) when ncat gets EOF
> > > > in input?
> > > >
> > > > I used following patch. It works, but I don't know if it's correct way and
> > > > correct place for this.
> > > >
> > > > --- nmap-6.01/ncat/ncat_connect.c.shutdown       2012-06-19
> > > > 09:49:19.000000000 +0200
> > > > +++ nmap-6.01/ncat/ncat_connect.c       2012-08-02 15:10:34.173929687 +0200
> > > > @@ -713,7 +713,7 @@ static void read_stdin_handler(nsock_poo
> > > >           if (o.sendonly) {
> > > >               /* In --send-only mode, exit after EOF on stdin. */
> > > >               nsock_loop_quit(nsp);
> > > > -        }
> > > > +        } else shutdown(nsi_getsd(cs.sock_nsi), SHUT_WR);
> > > >           return;
> > > >       } else if (status == NSE_STATUS_ERROR) {
> > > >           loguser("%s.\n", socket_strerror(nse_errorcode(evt)));
> > > >
> > > > Cheers,
> > > > Michal
> > >
> > > Sounds good to me. Any concern for supported non UNIX-like platforms?
> >
> > This function exists on UNIXes and Windows, I did not check other
> > platforms. Seems, there is one difference - windows uses different
> > defines, so this is needed:
> >
> > #ifndef SHUT_WR
> > #   define SHUT_WR SD_SEND
> > #endif
> >
> > in this file or in some header file.
>
> This sounds good to me. Henri, since you have already looked at this
> patch, would you apply it?
>
> Ideally there should be a new test for this in test/ncat-test.pl.
> Michal, would you take a look at that file and see if you can write a
> simple test? It will be along the lines of 1) Do a server_client_test,
> 2) write to $c_in then close $c_in, 3) check that you read EOF from
> $s_out.

Hi,

I found that the eof shutdown patch was incomplete, because it fixes 
only one part of two. Attached is second part (server side) of the fix. 
This time with a test. I know nothing about perl except it exists, but I 
used test for previous part as template.

Cheers,
Michal

Attachment: ncat-shutdown2.patch
Description:

_______________________________________________
Sent through the nmap-dev mailing list
http://cgi.insecure.org/mailman/listinfo/nmap-dev
Archived at http://seclists.org/nmap-dev/